Production 512: Compression Revisited – Again

Every time I sit down to write a column, I end up learning something new. I either discover a previously unknown (to me) nuance of the subject, or in researching a specific term or event, I’ll stumble onto something profoundly different to my understanding. In my 15+ years of writing for RAP, there have been quite a few topics that I’ve touched on multiple times, and every time I have written about those topics, I have learned something new. In that spirit, I have decided to write a new Beginner’s Guide To Compression this month.

Prod512 Logo 400pxI have no doubt that almost every reader will have a pretty solid understanding of what compression is and probably uses it all the time. Long-time readers might recall I did this topic just a couple of years ago, but this time there is one important difference, at least to my understanding. By breaking it down to the essentials, I sincerely hope that even the most seasoned pro can find a new insight. If I’m lucky, some will have a true “A-HA!” moment that will deepen their understanding of this essential tool of our trade. If you are just getting started in your professional production journey, I hope this will be even more valuable.

Why compression? Why not equalization or reverb? Well, I plan to do a beginner’s guide for those as well, but compression can really improve the impact of your work, possibly more than any of the other gizmos in our toolbox. It can also destroy it. Too little compression can weaken your work to the point of camouflaging it in the surrounding audio, giving it zero, or less than zero impact. Too much compression in the worst case can make your work illegible to the listener. It can also cause “listener fatigue,” particularly in longer pieces, again reducing the impact of your work.

We need to give you some vocabulary here so the following will make sense. I want to start with something I have, belatedly in my career, just figured out. (I know, I’m slow.)

dB - Decibels are a math headache for the casual audiophile, involving pressure, frequency, distance and even temperature. 0dB is silence. As you add decibels, the sound gets louder. A military cannon measures about 140dB at 1 meter. For the mathematically impaired, if you’re about 1 yard away from the muzzle when it fires, there is a very good chance your eardrums will rupture and bleed, impairing your hearing forever. (If you stand in front of the muzzle, you won’t have to worry about being deaf…you’ll be dead.) The band Golden Earring was once measured at 150dB at a concert in Germany. YIKES! The Occupational Safety and Health Administration (OSHA) recommends hearing protection for anything over 90-95dB. Here’s what I have only just realized: dB and VU are NOT interchangeable. They measure two entirely different things.

Gain - taking a signal and pushing it through any system, you can amplify (ADD) or attenuate (SUBTRACT) to or from the signal strength. It is a relative measurement given in your DAW as…

Volume Units - or VU, which measures signal strength. If you do not amplify or attenuate the signal, it is at…

Parity - expressed as ±0VU. 

0dB is silence…NO decibels…NO sound. Mathematically and practically speaking, negative decibels cannot be measured. Just understand that Volume Units do not measure loudness, they measure signal strength. Decibels are a measure of loudness, from 0dB to 150dB (or higher, I suppose). If you increase or attenuate the gain, you DO increase or decrease the Volume Units in the system and the decibels on your amplified output, but they are not interchangeable. Think of it this way: when it’s a signal in your system being mixed and manipulated, it’s VU, when it’s sound coming out of your speakers, it’s decibels which are controlled by whatever your amp is doing. The ONLY way that saying something is -20dB is by making it a relative comparison to where the level was previously – but technically speaking that is still incorrect. It might seem correct, but you’re comparing oranges and horse apples.

So to begin, let’s state exactly what compression is and is not. Compression is a method of reducing the dynamic range. It is NOT a way to make everything louder, although that’s certainly what it seems like. The dynamic range is the difference between the softest sound and the loudest sound in your track. The best example of dynamic range I can offer is classical music, specifically Tchaikovsky’s 1812 Overture. The reason is because the original score starts with a very pastoral sound of flutes and light strings. By the end, the score calls for several cannon to be fired in time to the music. (Now you know why I used a cannon in the definition.)

Remember, our DAWs do not measure decibels; they measure signal strength in Volume Units or VU. The loudest signals are at 0VU (parity). The softest signals are effectively at -100VU. A typical fader will drop from -60VU to infinity (∞) in the last couple of millimeters of the draw. If the softest sound is say -96VU and the loudest hits the top, the dynamic range is -96VU. This is a very simplified explanation. The accurate one requires a TON of math, but the results are the same…mostly.

A technical sidebar: most DAWs indicate up to 12VU above zero for the maximum, but in actuality, your system has been attenuated -12VU to start with. Digital signals, by design, cannot exceed 0VU, ever. That’s just the math-truth of ones and zeros.

The object of compression is to increase the gain of the soft signal and attenuate the gain of the loud signal, narrowing the dynamic range. It makes the signal thicker, not louder. If you adjust the output of your compressor upwards, the level gets back up to ±0VU, but the softer signals (those flutes and strings) get pushed up, along with the cannon, making the signal seem louder. The actual compression algorithm does NOT make everything louder. The reason it’s important to distinguish between dB and VU becomes obvious at this point. If the effective lowest signal strength is -100VU and the maximum is 0VU, the maximum dynamic range is -100VU. How do you account for a cannon at 140dB?

People often confuse limiters with compressors. While they might have very similar effect, the limiter does not compress the dynamic range, it simply prevents the loudest parts from exceeding a threshold, which you set. If you then increase the input gain on the limiter, the softer signal gets pushed up while the louder sounds are prevented from increasing beyond your threshold. The effect is similar, but on limiters, it’s really easy to push hard enough to distort the louder signals. PLUS, you have a lot less control over how the signal is shaped. Limiters offer only two main controls; Input/Threshold and Output. A compressor gives you a LOT more shaping parameters, which if properly used can keep the signal very clean, even when pushed harder than a limiter.

These are the basic controls of a good compressor:

Threshold - how high the signal must be before attenuation starts on the louder signals. The increase in gain for the softer signal is always there and is very much affected by the input and the…

Ratio - how much compression. If the compression ratio is 3:1, the signal will have to cross the threshold by +3VU for the output to increase by +1VU. Typically, the human voice should get about 2:1 as should drum tracks. Bass tracks will shine at 4:1 while electric guitars are good from 2:1 to 6:1. Your results may vary. These are just good starting points.

Attack - how quickly compression is applied. If your VO has a lot of sudden peaks up front, mainly plosives and fricatives, you need a fast attack. You can slow it down for a bus compressor, like your final mix.

Release - how soon after the signal drops below the threshold the compression stops. This can really affect the shape of your sound. A short release gives a lot more oomph, making it sound compressed; like the compression is working harder. A longer release will give a much more natural sound, ramping the gain down slowly instead of being abrupt.

Knee - how the compressor reacts to signals once the threshold is passed. Hard Knee settings squeeze the signal immediately, and Soft Knee squeezes over a longer period as the signal goes beyond the threshold. It’s subtly different from Attack, like fine-tuning.

Make-Up Gain - boosts the already compressed signal because this process often attenuates the overall signal, sometimes significantly. Remember that a 3:1 ratio drags the peaks down from ±3 to ±0VU. Make-Up gain adds to your signal strength without adding any additional noise.

Output - this final control allows you to boost or attenuate the final signal to suit your particular piece of production. This seems to be the same thing as Make-Up Gain, but it’s more of a brute control.

Now, let me give you a short list of rules. Just remember that MY rules are more like suggestions. If you don’t test the boundaries of these rules, you’re not doing your job, which is to be inventive and adventurous. 

1. Avoid compressing your production music. Strain to avoid compressing playlist music, or what your station is playing between your production pieces. Some poor schmuck spent an entire career and untold hours making the compression on that music perfect. Some of the music you use will be too ‘open and airy’ for your purposes, so apply compression with care. If you’re using playlist music, particularly in Country, Rock and CHR/AC formats, it takes almost no effort to completely screw up the music by adding compression. It’s been compressed, usually a LOT. But…you do you.

2. Avoid ‘double-compressing.’ Adding a second round of compression on any signal will almost guarantee you a major malfunction.

3. Use different compression set-ups on different voices. Every human voice reacts a little differently to compression, and it becomes very noticeable between genders and differently aged actors/announcers.

4. Most producers use just a touch of compression on the final mix. I do this, but I’ve been known to screw the pooch in this step. I once bounced the finished product to DropBox for my client to grab. Within minutes, I got a panicked email complaining about distortion. When I went back to the session, it sounded fine. THEN, I checked that last compressor on the output. OOPS! Always double-check your final mix.

Someone could probably write an entire book about audio compression, but I’ve… compressed it…down to just over 1800 words. {ahem} You’re welcome.

Comments (0)

There are no comments posted here yet

Leave your comments

  1. Posting comment as a guest. Your post will be moderated. Your email address will not be shown or linked. (If you have an account, log in for real time posting and other options.)
0 Characters
Attachments (0 / 3)
Share Your Location